1-bit Bonsai
2B-parameter vision-language model in 1-bit quantization for AI smart glasses on Snapdragon AR1.
Why it matters
Fits 4x more parameters in the same memory footprint for on-device visual reasoning.
What you could build with it
Run visual reasoning on-device for smart glasses and phones: receipt scanning and scene understanding with no cloud round-trip.
Does it hold up?
Convincing for compact on-device inference, but quality depends on PrismML's specialized quantization method - ordinary one-bit conversion is not equivalent.
Built with 1-bit Bonsai
- twinkites/bonsai-gardengithub · Garden of small browser-based LLMs running Bonsai 1-bit models entirely in-browser via WebGPU with a live demo.
- r13xr13/bonsai-harnessgithub · Third-party harness around Bonsai-8B with measured throughput: 368 tok/s on RTX 4090.
- PrismML-Eng/Bonsai-demogithub · 3,077 ★ · Official local demo: llama.cpp/MLX, Open WebUI, tool calling, vision, code interpreter and community hardware benchmarks.
- ThakiCloud/bonsai-1bit-reprogithub · 3 ★ · Independent reproduction: released weights genuinely work; naive binary quantization collapses; rotation and error-compensation recover much of the quality.
Learn more
First spotted on article: source.
More AI models
ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69Mistral Forge: enterprise platform for training and continuously improving proprietary modelsMistral AI launched Forge (Sep 28) — a full-lifecycle model training platform (pre-training, SFT, DPO/ODPO, RL…models · JEV 0.69kev: open, trainable Jev-like family of small decision models on Qwen3.5/3.8A family of small decision models (0.8B to 27B) built on Qwen3.5/3.8 that reproduces Jev's prefill-only typed-decision…models · JEV 0.68GPT-Synopsys: OpenAI and Synopsys multi-year deal to build an AI model that operates EDA toolsOpenAI and Synopsys signed a multi-year agreement on Sep 30 to jointly develop GPT-Synopsys, a specialized model that…models · JEV 0.66
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.